cannot reproduce Constant Crashing!

[jhayes_tucson]

I'm running PI 1.9.3 Lockhart, Build 1646 on a 16" MacBook Pro M1 (2021) running Monterey OS V12.4. I can't seem to run for more than about 1-2 hours before I get a segment violation error as shown in the attached screen shot. When this screen appears, I'm within anywhere from 1 hour to just a few operations before PI crashes and I lose everything. I'm forced to constantly save everything and the system often crashes as I'm trying to save my more recent work. I've uninstalled the code and done a clean reinstall numerous times to no avail. Is there any way to fix this or am I going to need to buy a Windows machine just to run PI? I am just about at my wits end with all the crashing! Does PI run any better on the newer Apple machines? I run multiple remote telescopes in Chile and PI has become the most unstable software that I use. This is super frustrating and it make PI almost unusable.

Thanks for whatever advice you can provide.

John Hayes

 

[Marcelofig]

May I ask, why are you still using Monterry, your Mac certainly supports the latest version of Mac OS Sequoia. That said, although Monterrey is not officially supported by PI, it shouldn't be causing that many problems.

More relevant, how much RAM does your Mac have and what camera are you using? It could be an issue of low memory vs large files, PI is notoriously RAM hungry.

And another question, are you using RC Astro XTs, have you updated them? I ask because there has been speculation around here that these could be the cause of the problem* and since their last update a couple of weeks ago I haven't seen this problem again on my Mac (mini M4 Pro, 64 GB RAM).

And come to think of it, there used to appear on this forum frequent complaints from Mac users reporting this problem, I myself experienced it a couple of times. But since a couple of weeks ago these have diminished.

(Now just for mentioning the lack of problems I am waiting for the next crash ).

The recommendation would be, with no guarantee of success, to upgrade Mac OS and update the XTs and see what happens.


* If you explore the forum you will see that every time this problem is reported by Mac users, the post has been tagged with the cannot reproduce tag. It is speculated that the cause of the problem is a third party process/script.

 

[WillB]

FWIW I have also been struggling with random crashes on a Mac M1 Studio Ultra (64GB RAM) for the last few years, currently the Mac is running Sequoia 15.4.1 but the Mac M1 Studio has been unstable running PixInsight for at least the previous two or three Mac O.S's too.

In fairness to the P.I. developers this is not a problem confined to P.I. alone as I also see random crashes with an expensive monthly subscription CAD program that supposedly runs on native Apple Silicon but the only way I can use that program on the Mac M1 is by running an old Mac Intel build of the application from several years back via Rosetta.

The common thread that links the two applications is the use of Qt.

Things I've found that help improve the stability of P.I. on the Mac Mi Studio Ultra.

1. Avoid using the Blink module as much as possible.
Blink will often crash P.I. when closing currently loaded images in the Blink module with a "Bus error" but you can reload the same images a second time and Blink will not crash, which is why I have not reported it because I can't find a set of images that will consistently reproduce the problem. As much as possible I use SubFrameSelector now in preference to Blink as SFS doesn't crash (so far).

2. Use a separate, standard, wired, two button and scroll wheel mouse when running PixInsight and forget about Apple's "Magic Mouse" or track pads with gesture support as Gesture Support on Magic Mouse seems to throw PixInsight into uncontrolled actions that can not be recovered from (such as uncontrolled zooming or desktop scrolling, doesn't crash P.I. but makes it unusable and only recovery option is a forced quit and restart.

3. Reduce the maximum number of processors used by at least one (Preferences > Parallel Processing and threads).
The M1 Ultra has twenty cores and I allow P.I. to use 18 so that when some other automatic Mac background process runs it doesn't have to fight for CPU space with P.I.

4. Create a dedicated PixInsight swap folder. I created one at the root of the user directory (Users/MyName/PixInsightTempSwap) and deleted the P.I. default folder at the /var/folders/ location (see point five and six below for the reason).

5. Create a desktop work folder specifically for PixInsight and load only copies of raw data to be worked on in that folder (see next point for reason).

6. If you have, or have ever had, Apples's TimeMachine configured on a Mac then you'll know that every hour TimeMachine will scan all folders and files on the Mac and create a backup. What most users don't know is that even if you are not connected to the TimeMachine backup disk TimeMachine will still create a compressed local backup "snapshot" to the internal HD every hour and with huge Gb's of image data constantly being written by some PixInsight process while TimeMachine is backing up that data will crash P.I. eventually, which seems to tie in with the oft quoted complaint that PixInsight on a Mac often crashes after an hour or so of use.

I have found that if you create a dedicated P.I. swap folder and delete the default one, plus create a dedicated work folder for PixInsight, and in TimeMachine setup specifically exclude those two folders from the TimeMachine backup then PixInsight is less likely to crash while TimeMachine is either writing to the real external backup disk or creating a local snapshot on the internal HD.

7. Lastly, after any P.I. crash, or when upgrading/downgrading PixInsight, or before starting a new project, with PixInsight closed, inspect the contents of the swap folder and clear any swap files that may still be there and in the normally hidden 'Macintosh HD/Users/YourName/Library/PixInsight' directory delete any files with the extension ".cache".

With the above few steps I can run most PixInsight projects to completion without a crash and I'm hoping that when Juan releases an Apple Silicon native version of PixInsight that it will be more stable but given my experience with my CAD application that runs natively on Apple Silicon I'm not counting on that as there do seem to be ongoing issues with Qt on the Apple Silicon Macs.

I suspect I'll end up buying a machine to run Linux and PixInsight alone and have been putting off the decision until Juan releases a Mac Silicon build, if that is still unstable I'll have no choice but to invest in a Linux machine.

Will.

 

[jhayes_tucson]

May I ask, why are you still using Monterry, your Mac certainly supports the latest version of Mac OS Sequoia. That said, although Monterrey is not officially supported by PI, it shouldn't be causing that many problems.

More relevant, how much RAM does your Mac have and what camera are you using? It could be an issue of low memory vs large files, PI is notoriously RAM hungry.

And another question, are you using RC Astro XTs, have you updated them? I ask because there has been speculation around here that these could be the cause of the problem* and since their last update a couple of weeks ago I haven't seen this problem again on my Mac (mini M4 Pro, 64 GB RAM).

And come to think of it, there used to appear on this forum frequent complaints from Mac users reporting this problem, I myself experienced it a couple of times. But since a couple of weeks ago these have diminished.

(Now just for mentioning the lack of problems I am waiting for the next crash ).

The recommendation would be, with no guarantee of success, to upgrade Mac OS and update the XTs and see what happens.


* If you explore the forum you will see that every time this problem is reported by Mac users, the post has been tagged with the cannot reproduce tag. It is speculated that the cause of the problem is a third party process/script.

Hi Marcelo,
Sorry for the slow reply. I've been traveling and now finally getting back to this thread.

My Mac has 64 GB of RAM with a 2 TB main SSD drive. I am running Monterey mainly because 1) I'm always using my Mac 2) it virtually never crashes, and 3) I absolutely despise updates. I'm was an early adopter in the PC world and I used to eagerly run the latest of everything; however, after having been burned countless times, I changed my ways. Now, once something works, I hate having to fix it. Updates often break things and add worthless bloated new features so I put it off until it is absolutely necessary. My earliest versions of PI NEVER crashed...and now, after many years of updates, I can't get through a single image without multiple restarts. If updating the OS would fix the problem, I'd do the update in a heartbeat but since that shouldn't be the problem, let's put that question aside--at least for now.

I recently checked and all of my RC Astro plugins are the latest versions. That's a good suggestion but unlikely to be causing causing my problems.

If I could reproduce the problem myself, I'd post it. Unfortunately, although it happens regularly, there doesn't seem to be a pattern to what causes PI to crash. Sometimes it will run for 2 days of intense work after the first appearance of the red "Segment Violation" error screen and sometimes it crashes within 10 minutes. Either way it is super frustrating.

Thanks for giving it some thought and for taking the time to respond. I really appreciate it!

John

 

[jhayes_tucson]

FWIW I have also been struggling with random crashes on a Mac M1 Studio Ultra (64GB RAM) for the last few years, currently the Mac is running Sequoia 15.4.1 but the Mac M1 Studio has been unstable running PixInsight for at least the previous two or three Mac O.S's too.

In fairness to the P.I. developers this is not a problem confined to P.I. alone as I also see random crashes with an expensive monthly subscription CAD program that supposedly runs on native Apple Silicon but the only way I can use that program on the Mac M1 is by running an old Mac Intel build of the application from several years back via Rosetta.

The common thread that links the two applications is the use of Qt.

Things I've found that help improve the stability of P.I. on the Mac Mi Studio Ultra.

1. Avoid using the Blink module as much as possible.
Blink will often crash P.I. when closing currently loaded images in the Blink module with a "Bus error" but you can reload the same images a second time and Blink will not crash, which is why I have not reported it because I can't find a set of images that will consistently reproduce the problem. As much as possible I use SubFrameSelector now in preference to Blink as SFS doesn't crash (so far).

2. Use a separate, standard, wired, two button and scroll wheel mouse when running PixInsight and forget about Apple's "Magic Mouse" or track pads with gesture support as Gesture Support on Magic Mouse seems to throw PixInsight into uncontrolled actions that can not be recovered from (such as uncontrolled zooming or desktop scrolling, doesn't crash P.I. but makes it unusable and only recovery option is a forced quit and restart.

3. Reduce the maximum number of processors used by at least one (Preferences > Parallel Processing and threads).
The M1 Ultra has twenty cores and I allow P.I. to use 18 so that when some other automatic Mac background process runs it doesn't have to fight for CPU space with P.I.

4. Create a dedicated PixInsight swap folder. I created one at the root of the user directory (Users/MyName/PixInsightTempSwap) and deleted the P.I. default folder at the /var/folders/ location (see point five and six below for the reason).

5. Create a desktop work folder specifically for PixInsight and load only copies of raw data to be worked on in that folder (see next point for reason).

6. If you have, or have ever had, Apples's TimeMachine configured on a Mac then you'll know that every hour TimeMachine will scan all folders and files on the Mac and create a backup. What most users don't know is that even if you are not connected to the TimeMachine backup disk TimeMachine will still create a compressed local backup "snapshot" to the internal HD every hour and with huge Gb's of image data constantly being written by some PixInsight process while TimeMachine is backing up that data will crash P.I. eventually, which seems to tie in with the oft quoted complaint that PixInsight on a Mac often crashes after an hour or so of use.

I have found that if you create a dedicated P.I. swap folder and delete the default one, plus create a dedicated work folder for PixInsight, and in TimeMachine setup specifically exclude those two folders from the TimeMachine backup then PixInsight is less likely to crash while TimeMachine is either writing to the real external backup disk or creating a local snapshot on the internal HD.

7. Lastly, after any P.I. crash, or when upgrading/downgrading PixInsight, or before starting a new project, with PixInsight closed, inspect the contents of the swap folder and clear any swap files that may still be there and in the normally hidden 'Macintosh HD/Users/YourName/Library/PixInsight' directory delete any files with the extension ".cache".

With the above few steps I can run most PixInsight projects to completion without a crash and I'm hoping that when Juan releases an Apple Silicon native version of PixInsight that it will be more stable but given my experience with my CAD application that runs natively on Apple Silicon I'm not counting on that as there do seem to be ongoing issues with Qt on the Apple Silicon Macs.

I suspect I'll end up buying a machine to run Linux and PixInsight alone and have been putting off the decision until Juan releases a Mac Silicon build, if that is still unstable I'll have no choice but to invest in a Linux machine.

Will.

Thanks for your thoughts Will. My Mac all on it's own, running my super old OS is amazingly stable. I might only have to reboot it every year or so and that's usually only after I've totally overloaded the memory with data and apps all running concurrently. In order of your comments:

1) I've never had a problem with Blink but I'm also careful about how many images I load. Regardless, I also normally use SFS and WBPP. I had not even used Blink the last time that PI crashed.
2) I use a two button, Bluetooth Logitech mouse. The Apple "Magic Mouse" is junk.
3) Reducing the number of processors is an interesting suggestion and I'll try it.
4) Another interesting suggestion that I'll try!
5) I set up a dedicated swap folder a number of updates ago but I forgot all about that so I'll have to go back to check how it is now configured.
6) I gave up on Time Machine about 10 years ago. My machine backs up to the cloud.
7) Ok...that's another good thing to try.

Hopefully the move to native Apple ARM code will fix these problems! That would certainly be an update that I would immediately jump all over. Of course, that also the kind of update that could also be riddled with its own problems; but, we'll see.

Again...thanks a million for your suggestions!!

John

 

[Marcelofig]

I recently checked and all of my RC Astro plugins are the latest versions. That's a good suggestion but unlikely to be causing causing my problems.

If I could reproduce the problem myself, I'd post it. Unfortunately, although it happens regularly, there doesn't seem to be a pattern to what causes PI to crash. Sometimes it will run for 2 days of intense work after the first appearance of the red "Segment Violation" error screen and sometimes it crashes within 10 minutes. Either way it is super frustrating.

Thanks for giving it some thought and for taking the time to respond. I really appreciate it!

John

Yes the problem, as you can see from the tag on this post, is that the PI team has not been able to reproduce it.

There is a post by Juan, which I can't find now, in which he describes how they perform their tests, it is an almost clinical environment, without any third party installed and I think even without internet connection. Under this environment they have not been able to reproduce the problem, which leads us to believe that the problem is caused by some third party process and XTs are a very popular process.

Now I haven't experienced it myself recently (Mac mini M4 Pro 64 GB RAM with Mas OS Sequoia), but I'm far from being a heavy user. The hope for Mac users is that at some point there will be a native version of PI for ARM chip and this will solve the problems.


PS: No problem with the delay in responding, hopefully you've been setting up some new telescope in some remote part of the world.

 

[pfile]

6) I gave up on Time Machine about 10 years ago. My machine backs up to the cloud.

this can still cause problems with PI, apparently when whatever cloud backup daemon is running and happens to open files that PI is also trying to open. so it might be a good idea to exclude all PI disk areas from your cloud backup even if you aren't using TimeMachine.

but also i'm not sure you can stop the APFS snapshots, i think those might happen independently from TM.

i think your strategy is sound, i always hang back one or two versions of OSX. but at some point apple stops supporting the OS and then security bugs pile up to the point where it might not be safe using an old version OSX connected to the internet. so i always upgrade to the n-1 or n-2 version of OSX when support ends for the one i am using.

rob